Galactic Swarm-based Harris Hawks Optimization for Handling the Upstream/Downstream Configuration in PON-SDN-DSVM Systems: A Deep Learning Concept

Abstract

Abstract Owing to the strategic benefits, recent broadband ONs have attracted the interest of academics and business (cost-efficiency, flexibility, vast bandwidth, and mobility). Simultaneously, the rise of SDN allows for effective dynamic reconfiguration of fundamental network elements via SDN controllers. As a result, efficient traffic-aware strategies for automatically identifying appropriate specifications to improve network performance are possible. In order to do this, a new deep learning technique for an SDN-defined hybrid PON is suggested. A PON-SDN-DSVM makes up the proposed design. The introduced hybrid deep learning model called DSVM is formed by combining the DNN and SVM and it is optimized by the integration of two well performing algorithms such as GSO and HHO in order to produce the hybrid meta heuristic algorithm called as GS-HHO with the consideration of accuracy maximization. The suggested technique obtains traffic-aware knowledge from SDN controllers and adjusts the communication's upstream-downstream configuration. On the basis of the traffic dynamics of the complete hybrid network, this traffic-aware method can determine the best configuration. Real-world traffic traces are used to test the proposed method in a realistic context. The suggested technique, according to the attained numerical findings, provides considerable increases in network performance with respect to accuracy, precision, f1 score, recall, comparison, and confusion matrix analysis.